In the United States, the issue of whether “natural immunity” from prior infection is as good as vaccination is becoming politicised.
Fortunately, the Office for National Statistics has released results from a sophisticated analysis of its Covid-19 infection survey that address this issue. Helped by test-and-trace data, it identified respondents’ “infection episodes” and used other data to infer whether participants were previously vaccinated, had a previous infection or neither.
Analysts used “Poisson regression” to estimate the risk of an infection episode, adjusting for known risk factors such as age and deprivation. Rather than counting people, the study considered days each person was at risk of infection. All risks are relative to having no previous vaccination or recorded infection. The analysis looked at two periods. Between December 2020 and mid-May 2021, when the Alpha variant was dominant, full vaccination reduced the risk of testing positive by an estimated 79%, with little difference between vaccines. Previous infection, without vaccination, reduced that risk by around 65%. The modelled uncertainty around each estimate was roughly six points: vaccination seemed more protective than prior infection in this period.
Between mid-May and mid-August 2021, when the Delta variant reigned supreme, full vaccination reduced the risk of testing positive by 64-70%. Past infection without vaccination had a similar effect (65-77%). Across all analyses, two vaccine doses were more effective than one. There was higher protection against getting Covid-19 with symptoms.
This is a challenging question. Deliberate infection has ethical problems so a randomised trial can’t be performed. This study demonstrates the power of a large and expensive survey, supplemented by linkage to other data sources but, even using regression methods, we cannot be sure our groups are fully comparable. For example, some groups may differ in their tendency to get tested. Other analyses look at regularly tested populations such as healthcare workers.
The results may add to the intensity of the debate over mandated vaccination, but it is vital that policy decisions are based on the best possible statistical analysis.